Brain imaging reliably reveals anatomy and broad patterns of which regions are relatively more active during a task, but it does not read specific thoughts, reliably detect lies, or prove that a region causes a behaviour. The phrase a region lit up is loose shorthand for a statistical difference in a slow, indirect signal. Most of the myths around imaging come from three predictable errors: treating activation as mind-reading, reasoning backwards from a region to a mental state, and mistaking correlation for cause.
What imaging genuinely shows
Start with the real power, because it is considerable and easy to lose sight of amid the caveats. Structural imaging shows the anatomy of a living brain with remarkable clarity, and it does so well enough to transform how tumours, strokes, and tissue changes are found and followed. That is not a modest achievement dressed up; it is one of the most useful diagnostic advances of the last century, and it is not in serious dispute.
Functional imaging, used carefully, reliably reveals broad organisation. It has confirmed, across thousands of studies, that particular kinds of processing tend to engage particular networks: vision draws heavily on the back of the brain, movement on a strip near the top, language on regions usually toward the left. It shows that the brain is organised into networks that coordinate, that these networks are active even at rest, and that their broad layout is fairly consistent across people. These are genuine, replicated findings. The honest summary is that imaging is excellent at where in a coarse, group-averaged sense, and it has mapped the rough functional geography of the brain in real detail.
Imaging is a superb map-maker and a poor mind-reader. It can tell you the neighbourhood where something happened; it cannot tell you the private thought that happened there.
What it can and cannot show
Setting the strengths and limits side by side is the fastest way to calibrate what a scan means when you meet one in the wild.
What it can show
The brain's anatomy in fine detail. Which broad networks tend to engage during a class of task, averaged across people. Reliable differences between groups on average, for example that a network is on the whole less active in one group than another. The rough functional geography of vision, movement, language, and attention. Changes over time in the same person, such as recovery after injury.
What it cannot show
The content of a specific thought, memory, or intention. Whether a single person is lying, in any trustworthy way. A one-to-one map from a region to a single mental state. That a region caused a behaviour rather than merely accompanying it. A diagnosis of most conditions in an individual from a scan alone. Fine-grained timing, given the slowness of the blood signal.
The three myths, and why they are wrong
Almost every misleading claim about brain imaging is a version of one of these three errors. Naming them makes bad coverage easy to spot.
A brain scan can read your mind or reliably catch you lying.
Functional imaging measures a slow, averaged, indirect signal, not the content of thought. Attempts at brain-based lie detection have not proven reliable enough to trust, particularly against people trying to beat them, and courts have generally refused to admit them. A scan can show a region was relatively busier; it cannot pull out the sentence you were thinking or certify that you told the truth.
If a region linked to an emotion lit up, the person must have felt that emotion.
This is reverse inference, and it is a genuine logical fallacy. Most brain regions take part in many different functions, so a region often associated with fear also does plenty of other jobs. Seeing it active does not establish that the person felt fear, any more than seeing a busy motorway proves everyone on it is going to the same place. Researchers such as Russell Poldrack have shown how often this error slips into confident claims.
A region lighting up during a task proves that region causes the behaviour.
Imaging shows correlation: a region was more active while something happened. Correlation is not cause. The activity could be a side effect, a downstream consequence, or a region simply keeping company with the one that matters. Establishing cause needs other methods, such as watching what fails when a region is temporarily disrupted or lost to injury. The tidy phrase lit up quietly hides all of this.
Why the loose language misleads
A lot of the trouble comes from three innocent-sounding shorthands. Lit up suggests a light switching on, when the truth is a small percentage change in a blood-oxygen signal that survived a statistical threshold. A region for a function suggests one region does one job, when regions are multitaskers and functions are spread across networks. And a scan of the brain suggests a photograph, when it is a reconstruction with processing choices baked in, as the overview explains. None of these phrases is a lie, but each nudges a careful finding toward a dramatic story.
A quick test for any imaging claim. Ask three things. First, is this about a group average or about one individual? Scans describe groups far better than persons. Second, is it reasoning forwards from a task to activity, which is fair, or backwards from activity to a mental state, which is reverse inference and often unsound? Third, is the claim about correlation or cause, and if cause, is there evidence beyond the scan itself? A claim that survives all three questions is worth taking seriously. One that fails them is probably a colourful picture stretched past what it can bear.
Holding both truths at once
The point of all this caution is not to dismiss brain imaging. It is to read it accurately, which means holding two truths together. Imaging really did let us see the living brain and map its rough functional geography, and that is a lasting achievement. At the same time, it is an indirect, statistical, group-level tool that cannot read minds, detect lies, or prove cause on its own. Both statements are true, and the honest reader keeps them in the same frame. The research page carries this forward, sorting specific claims into what is settled, what is mixed, and what remains genuinely contested.
Why decoding is not mind-reading
One genuine advance deserves careful handling, because it is the finding most often stretched into the mind-reading myth. In tightly controlled experiments, researchers can sometimes train software to tell, from a person's brain activity, which of a small, known set of pictures or categories they are currently viewing. This is real, and it is impressive. But it is worth being precise about what it is and is not. The software is trained in advance on that specific person looking at that specific set of options, and it chooses among those pre-defined possibilities. It is closer to a very sophisticated multiple-choice guess than to eavesdropping on free thought.
The distance from there to reading an open-ended, spontaneous thought is enormous. Take away the pre-training, the constrained choices, or the same person's own calibration data, and performance falls off sharply. Nobody can point a scanner at a stranger and print out the sentence in their head, and there is no evidence the current approach scales to that. So when a headline says scientists decoded someone's thoughts, the accurate translation is usually that software distinguished a handful of pre-selected options above chance in a controlled setting. That is a real result worth celebrating on its own terms, and it is not mind-reading.
Where to go next
To weigh the evidence claim by claim, read the research page. To revisit why the signal is indirect in the first place, see how brain imaging works. Or return to the overview for the big picture.
Sources
- Poldrack RA. Can cognitive processes be inferred from neuroimaging data? Trends in Cognitive Sciences. 2006;10(2):59-63.
- Logothetis NK. What we can do and what we cannot do with fMRI. Nature. 2008;453:869-878.
- Farah MJ, Hutchinson JB, Phelps EA, Wagner AD. Functional MRI-based lie detection: scientific and societal challenges. Nature Reviews Neuroscience. 2014;15:123-131.
This page is educational and is about interpreting imaging findings, not diagnosing any individual. The activation described here is a statistical, group-level pattern in an indirect signal, not a direct readout of thought.